AI-Supported Audit Automation

At the 14th Trend Conference on Banking Supervision organised by msg for banking ag — with around 100 participants on-site and online at the Stadthaus am Markt in Frankfurt — Prof. Dr. Dirk Schieborn and Kay Dietrich (Head of Internal Audit Business, Aareal Bank) presented their joint project experience report under the title "AI@AUDIT: AI-Supported Audit Automation".

The central question was how generative AI can fundamentally transform operational audit work in a bank's internal audit — not as a replacement for the auditor, but as an intellectual exoskeleton that handles routine work and frees up capacity for contextual judgement.

The talk was structured in three parts: First, a maturity model for LLM adoption in audit was introduced — from simple chatbot to RAG systems to fully integrated workflow automation (Level 3: "Away from chat, towards process"). The second part presented the AI@AUDIT client project at Aareal Bank: the AI autonomously analyses contracts, checks every clause against DORA/EBA requirements, and delivers structured finding drafts with source references and reasoning — while the auditor focuses on strategic assessment and final evaluation. A three-stage quality assurance methodology ("Engineered Trust") — comprising ground-truth validation, human-in-the-loop curation, and stability testing — ensures the reliability of AI outputs.

The third part introduced the research project auditissimo.com — a joint initiative with Reutlingen University and Steinbeis Analytics. The goal is a validated AI suite to support IRBA model audits: six integrated modules along the audit lifecycle (from the regulatory knowledge base through gap analysis and deep-dive review to automated reporting), integrating GenAI seamlessly into the entire audit process.

The conference also covered current developments on MaRisk 2026 (Dr. Torsten Kelp, BaFin), the Omnibus Package and ESG requirements, as well as CSRBB and SREP 2.0.